AI Uncovers Six Secret Zones Hiding at the Bottom of Earth's Interior

A Discovery Hidden in Plain Sight for Decades

Over two million waveforms from roughly 5,000 magnitude-6-plus earthquakes had been sitting in seismic archives for decades. Manual analysis had picked through this archive repeatedly — and kept missing something. A deep-learning system developed by researchers at the Chinese Academy of Sciences didn't. The result is one of the most significant discoveries in deep-Earth science in years: six previously unknown structural zones lurking near the boundary between Earth's mantle and its outer core.

Published in the Journal of Geophysical Research: Solid Earth, the deep-learning system identified 174,929 faint seismic signals known as PKP precursors across nearly 35 years of seismic data. That number is staggering. The system identified 174,929 high-quality PKP precursor signals — more than ten times all previous studies combined. In other words, everything scientists had ever catalogued before this AI came along was just a fraction of what was actually there.

What Are PKP Precursors — and Why Do They Matter?

The signals in question are called PKP precursors: faint echoes that scatter off irregularities where solid mantle rock meets the liquid iron-nickel outer core, roughly 2,900 km down. They arrive just before the main seismic wave, weak enough to drown in noise. Finding them manually was slow, subjective, and geographically patchy. Think of it like trying to hear a whisper in a stadium — the signal is real, but the surrounding noise makes it nearly impossible to detect without help.

The researchers trained a deep-learning system to do the initial screening, which enabled them to examine millions of seismic waveforms from almost 5,000 earthquakes between 1990 and 2024. The algorithm first sorted recordings by quality, then determined whether they contained PKP precursors. Crucially, the researchers manually checked and corrected the AI model's mistakes, then fed the corrected examples back into the algorithm during the training process. It was a collaboration — human expertise guiding machine scale.

Six New Zones, and Some Tantalizing Theories

The six new zones, designated B1 through B6, are located near the boundary between Earth's solid mantle and liquid outer core, approximately 2,900 kilometres under the surface. These newly charted zones sit beneath high-latitude locations, including parts of Eurasia, Central Asia, and the South Atlantic Ocean. And unlike what previous research suggested, these aren't random isolated blips. The breakthrough reveals continuous belts of deep material rather than isolated anomalies, providing new exploration targets for geophysicists.

The newly identified zones mark regions where temperature and composition differ sharply from the surrounding mantle material. Researchers suggest that these deep-seated scatterers may originate from ancient subducted slab remnants dragged down over billions of years, localized partial melting, or mineral phase transitions. One especially dramatic hypothesis has also circulated: some coverage has floated the idea that these zones could contain material from Theia, the Mars-sized object that collided with early Earth to form the Moon — though that remains squarely in hypothesis territory.

Why This Matters Far Beyond the Lab

Understanding the core-mantle boundary carries direct implications for surface dynamics. The boundary region dictates the transfer of heat from Earth's interior to its surface — a thermal engine that drives volcanic activity, the formation of mountain ranges, and the movement of tectonic plates. Heat flow at this boundary also drives Earth's geodynamo, the mechanism behind the magnetic field protecting us from solar radiation. In short, what happens nearly 3,000 kilometers below your feet shapes the world you live on.

The PKP precursor catalogue will keep growing. Researchers expect it to sharpen fine-scale models of the lowermost mantle and eventually enable similar methods on other planets, once sufficient seismic data exists. The broader lesson here may be just as important as the discovery itself: vast archives of scientific data, collected over generations, still hold secrets — and AI is proving to be the key to unlocking them.